• Skip Baidu live fetch in the scheduled config verification to prevent CI timeouts.
• Keep Baidu selectors/adapters intact while preserving live coverage for other engines.
• Update verifier documentation to match the enabled default checks.

graph TD
  A["Scheduled/Manual verify job"] --> B["verify-search-engine-configs.mjs"] --> C{"URL list"}
  C --> D["Fetch HTML"] --> E["Site adapters"] --> F["Pass/Fail report"]
  C --> G["Baidu URL disabled"]
  subgraph Legend
    direction LR
    _job["CI job"] ~~~ _script["Verifier script"] ~~~ _decision{"Config"}
  end

Loading

The following are alternative approaches to this PR:

1. Skip Baidu only on GitHub-hosted runners

• ➕ Preserves Baidu live coverage for local/self-hosted environments
• ➕ Keeps scheduled CI stable without permanently removing coverage everywhere
• ➖ Adds environment detection/branching logic
• ➖ Still leaves the main CI signal without Baidu coverage (where it matters most)

2. Add strict per-request timeouts + abort/retry

• ➕ Could keep Baidu enabled while preventing 6-hour hangs
• ➕ General hardening for other engines if they become flaky
• ➖ Doesn't address frequent connection resets; could still be noisy red
• ➖ More moving parts (timeouts/backoff) to tune and maintain

Recommendation: The PR’s approach (removing Baidu from default live URLs while keeping its selector/adapter configuration) is the best near-term fix because it restores CI reliability with minimal change surface. If Baidu live coverage is still desired, consider re-enabling it conditionally (e.g., non-GitHub runners) plus adding request timeouts to prevent long hangs.

Bug fix (1) +5 / -2

verify-search-engine-configs.mjsDisable Baidu live verification URL to prevent flaky CI fetches +5/-2

Disable Baidu live verification URL to prevent flaky CI fetches

• Removes the Baidu search URL from the default live URL set by commenting it out and documenting the GitHub Actions connectivity issues. Keeps the rest of the engine URL configuration intact and slightly generalizes the Accept-Language comment.

.github/workflows/scripts/verify-search-engine-configs.mjs

Documentation (1) +3 / -3

AGENTS.mdAlign verifier docs with default engines (remove Baidu from defaults) +3/-3

Align verifier docs with default engines (remove Baidu from defaults)

• Updates the 'npm run verify' documentation to reflect that default checks target Bing, Yahoo, and Naver (no Baidu). Moves Baidu into the optional engines list and updates the default URL examples accordingly.

AGENTS.md

Walkthrough

The search verification script now uses site-specific fetch implementations, request timeouts, same-origin redirect checks for Baidu, non-200 response rejection, and coded error reporting. The GitHub Actions verification job has a five-minute execution limit.

Changes

Search verification

Estimated code review effort: 2 (Simple) | ~10 minutes

Sequence Diagram(s)

sequenceDiagram
  participant VerifyConfigs
  participant Verifier
  participant SearchEngine
  VerifyConfigs->>Verifier: Start verification job
  Verifier->>SearchEngine: Fetch page with timeout
  SearchEngine-->>Verifier: Return redirect or HTTP response
  Verifier->>Verifier: Validate origin, status, and response text
  Verifier-->>VerifyConfigs: Report result and error code

• X
• Mastodon
• Reddit
• LinkedIn

• .github/workflows/scripts/verify-search-engine-configs.mjs
• .github/workflows/verify-configs.yml

• Baidu coverage removed (file-level, prior state): no longer applicable — HEAD keeps Baidu live via globalThis.fetch.
• Mixed Headers instances (line 150, Copilot): resolved — desktopHeaders is now a plain object and the node-fetch Headers import was removed.
• Skip Baidu rather than switch fetchers (line 148, codex): author provided justification; Baidu live coverage intentionally retained.

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 9129b17)

Status: No Issues Found | Recommendation: Merge

Previous review (commit 7fbd1df)

Status: No Issues Found | Recommendation: Merge

Description

Baidu is no longer verified because its URL list is now empty, and the verifier treats an empty list
as success (no fetches, no selector assertions, no skip notice). Since Baidu remains a runtime site
adapter, selector/parser regressions for Baidu can now land without any verification signal.

Code

.github/workflows/scripts/verify-search-engine-configs.mjs[R100-103]

+  baidu: [
+    /*'https://www.baidu.com/s?wd=hello'*/
+    // GitHub Actions requests are reset or left hanging
+  ],

Evidence

The verifier only validates selectors inside urlArray.map(...), so an empty baidu URL array
results in zero fetches/DOM parses for Baidu and no failures being recorded; meanwhile Baidu is
still configured and initialized as a runtime adapter.

.github/workflows/scripts/verify-search-engine-configs.mjs[90-114]
.github/workflows/scripts/verify-search-engine-configs.mjs[146-185]
src/content-script/site-adapters/index.mjs[32-80]
src/content-script/site-adapters/baidu/index.mjs[1-26]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The verifier currently skips Baidu implicitly by using an empty URL array, which removes verification coverage without reporting that Baidu was skipped.

## Issue Context
Baidu is still present in the shared `config` and is mounted by the runtime site adapter, so losing verification coverage can allow regressions to ship undetected.

## Fix Focus Areas
- .github/workflows/scripts/verify-search-engine-configs.mjs[90-114]
- .github/workflows/scripts/verify-search-engine-configs.mjs[146-185]

## Suggested fix
- Add an explicit skip mechanism (e.g., `SKIP_ENGINES=baidu` or default-skip Baidu when `process.env.GITHUB_ACTIONS` is set) rather than relying on an empty array.
- Log skipped engines (e.g., `console.log('Skipping baidu: disabled in CI')`) so the verifier output reflects reduced coverage.
- Optionally add a fixture-based Baidu check (parse a stored HTML snapshot with JSDOM) so selectors are still validated without making network calls.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools